Superfluids, vortices and spinning charged operators in 4d CFT

We include vortices in the superfluid EFT for four dimensional CFTs at large global charge. Using the state-operator correspondence, vortices are mapped to charged operators with large spin and we compute their scaling dimensions. Different regimes are identified: phonons, vortex rings, Kelvin waves, and vortex crystals. We also compute correlators with a Noether current insertion in between vortex states. Results for the scaling dimensions of traceless symmetric operators are given in arbitrary spacetime dimensions. ∗gabriel.cuomo@epfl.ch ar X iv :1 90 6. 07 28 3v 2 [ he pth ] 2 2 Ju n 20 19


Summary of results
Let us first set our conventions for the four dimensional rotation group SO (4). Spinning operators in four dimensions are classified in representations labelled by two positive half-integer quantum numbers (J,J). These are related to the maximal values allowed for the Cartan generators J 34 and J 12 as With no loss of generality, we assume J 34 ≥ J 12 ≥ 0. The main prediction of the superfluid EFT is the scaling dimension of the lightest scalar operator of charge Q in the spectrum. It is given by [13] ∆ 0 (Q) = αQ 4/3 + βQ 2/3 + . . . , (1.2) for Q 1; here α and β are independent Wilson coefficients. In this work, we compute the scaling dimension of the lightest operator as the spin is increased. As in [30], the EFT describes the regime where the spin is below the unitarity bound, J,J Q 4/3 , and cannot reach the regime analyzed by the analytic bootstrap [7,8,31,32,33,34,35,36,37,38,39]. To leading order in the charge and the spin, the results depend on the first coefficient in (1.2) and on an extra coefficientγ parametrizing the vortex tension.
• For 2 ≤ J 12 ≤ J 34 Q 1/3 the minimal energy state is given by two phonons propagating on the superfluid, with energy: • For 1 ≤ Q − J 34 Q and 2 ≤ J 12 Q 1/3 , the lowest energy state corresponds to a Kelvin wave of spin J 12 propagating on a large vortex ring. The corresponding operator scaling dimension is given by: (1.11) These results apply to CFTs whose large charge sector can be described as a superfluid and which admit vortices. These are natural and simple conditions, hence we expect them to apply to a wide range of theories with a U (1) symmetry. Nonetheless, we cannot prove these assumptions. The rest of the paper is organized as follows. In section 2 we review the superfluid description of large charge operators as well as the vortex EFT in 2+1 dimensions. In 3 we formulate the effective field theory (EFT) for vortices in 3 + 1 dimensions. The results of this section are derived in sec. 4. In 5 we show how to make predictions for correlators involving a current insertion between two vortex states and in 6 we briefly comment on how the results (1.4) and (1.6) change in generic dimensions. Finally in 7 we draw our conclusions and comment on future research directions. Technical details are given in the appendices A, B, C.
2 Review of previous results

Conformal superfluid
Let us first remind that, in a CFT, the state-operator correspondence relates eigenstates of the Hamiltonian H on S d with the set of local operators at any given point [11,12]. The quantum numbers of the state on S d and the corresponding operator are the same. In particular, the energy E is related to the scaling dimension of the latter as ∆ = E/R. The EFT description of CFTs at large quantum numbers is based on the assumption that the lightest scalar operator with U (1) charge Q in a d + 1 dimensional CFT corresponds to a state with homogeneous charge density on R × S d . For Q 1, the scale associated with the density of this state is parametrically bigger than the S d radius R and the CFT is expected to be in a "condensed matter phase". As argued in [13], the simplest possibility is that the CFT enters a superfluid phase. Technically, this is equivalent to assuming an effective description in terms of a U (1) Goldstone boson [21]. The effective Lagrangian is fixed by shift symmetry and Weyl invariance: We use the notation (∂χ) = (∂ µ χ∂ µ χ) 1/2 and c, c 1 , c 2 are Wilson coefficients. Here R µ νρσ is the Riemann tensor on the cylinder R × S d . We assume c, c 1 , c 2 ∼ O(1), corresponding to the generic expectation for a strongly coupled system. On a homogeneous background at finite charge, the field takes the value χ = µt, where µ is the chemical potential of the system. To leading order in derivatives, it is related to the U (1) charge density j 0 as where Ω d = 2π is the S d volume. The chemical potential sets the cutoff of the EFT: By the state/operator correspondence, the ground state of (2.1) corresponds to the minimal energy state with charge Q. Its energy is determined by a semiclassical analysis and takes the form: Quantum corrections provide a Q 0 contribution. For even d, there is no local counterterm correcting this term, which is hence universal [10]. The Lagrangian (2.1) describes also excitations on the background. For instance, expanding χ = µt + π and working to leading order in derivatives we get Quantizing the system, it follows that the spectrum can be organized as a Fock space in terms of single particle states with angular momentum J and energy given by are the eigenvalues of the Laplacian on S d and the sound speed c s is fixed by conformal invariance. Physically, these states correspond to phonons propagating in the superfluid and are associated with new operators in the CFT. The J = 1 mode has ω 1 = 1/R and corresponds to the creation of a descendant. A natural question is how the spectrum changes, as the spin J is increased. When the angular momentum is parametrically smaller than the cutoff (2.3), the spectrum is reliably described by phonons (2.6). The results (1.3) and (1.7) then follow. Increasing spin, one finds singular solutions with a non zero winding number, such as π = φ, where φ is the azimuthal angle. This corresponds to the fact that, for J Q 1/d , vortices develop in the superfluid and must be included in the effective description. This was done in [30] for a 2 + 1 dimensional CFT. The main goal of this work is to carry a similar analysis for a 3 + 1 dimensional conformal field theory. To build some intuition, we briefly review the results of the 2 + 1 dimensional EFT in the next section.

Vortices in 2+1 dimensions
In d = 2, the action (2.1) reads To study vortices, it is convenient to consider a dual description in terms of a gauge field. To this aim, we introduce an independent variable v µ ≡ ∂ µ χ and a Lagrange multiplier A µ to set the curl of v µ to zero: where µνρ / √ g is the antisymmetric Levi-Civita tensor. Integrating out v µ we get The U (1) current relates the two descriptions (2.10) As a consequence, the charge density (2.2) translates into a homogeneous magnetic field F θφ = B sin θ = Q 2R 2 sin θ, which sets the cutoff of the EFT according to (2.3). The action (2.9) describes a propagating degree of freedom, given by the fluctuations of the magnetic field F θφ and which corresponds to the phonon in the original picture, together with a nonpropagating Coulomb field A 0 , which does not have any local analogue in the scalar formulation. As we will see, it is precisely this extra component which provides the leading coupling to the vortices.
In the EFT, vortices are heavy charged particles in the dual description (2.9). They are treated as 0 + 1 dimensional worldlines, whose spacetime trajectory is parametrized by a function X µ p (τ ) of a time parameter τ . The action of the superfluid plus vortices is fixed by the requirement of Weyl invariance and τ -reparametrization invariance; the lowest orders in derivatives take the form [30] The second term is the minimal coupling between the gauge field and a particle of charge q p ; this cannot be written in a local form in the scalar picture, showing the convenience of the gauge formulation. Notice that the charge q p corresponds to the Goldstone winding number and is hence quantized: q p ∈ Z. The third term is the action for a relativistic point particle in a superfluid 1 ; it is multiplied by an arbitrary function of jµẊ µ jẊ , since the superfluid velocity breaks Lorentz symmetry and provides an alternative condensed matter metric [26]. Working in the physical gauge X 0 p = τ , we notice that the leading term in time derivatives for the vortex lines arises from the second piece in (2.11). As we will self-consistently see, this implies that vortices move with non-relativistic velocities |˙ X| ∼ 1/ √ B. Hence we can neglect terms with two time derivatives in the last term, retaining only a constant contribution proportional to √ B which is interpreted as the vortex mass. This procedure is sometimes called lowest Landau level approximation in the literature [40,41,42,43,44].
The equations of motion (EOMs) deriving from (2.11) are where E i = F i0 is the electric field and e 2 = 2 1/4 √ B 3κ . The particle EOMs (2.13) are first order in derivatives and imply that vortices move with drift velocity |˙ X p | ∼ | E/B| ∼ 1/ √ Q as anticipated. Consequently, particle velocities, as well as the magnetic field fluctuations sourced by them, are negligible and the only relevant interaction is the electrostatic one. We now look for static classical solutions of the EOMs (2.12) and (2.13). By the state/operator correspondence, these classical solutions will be associated to operators with the same quantum numbers. The spin and the scaling dimension of the corresponding operators are then determined classically from the energy momentum tensor. The scaling dimension of a state with n vortices reads 2 where X a p = (sin θ p cos φ p , sin θ p sin φ p , cos θ p ) is the vortex coordinate in the R 3 embedding of S 2 and ∆X 2 (x p , x r ) = 3 a=1 X a p − X a r 2 is the chordal distance between two vortices. The first term in (2.14) is the energy of the homogeneous phase, given by (2.4) with d = 2: The second term is the energy stored in the electric field sourced by the vortices, which is further rewritten as a sum over pairwise contributions in the right-hand side; the log Q ∼ log Λ 2 contribution arises from the logarithmically divergent self-energy of the point charges. We also used that the Figure 1: A vortex-antivortex pair moving on the sphere at fixed distance; in the stereographic projection the motion corresponds to two circular orbits.
net charge on the sphere must be zero p q p = 0, as a consequence of Gauss law. Finally, the last term is the contribution of n vortex masses and is written in terms of an independent coefficientγ, assumed to be the same for all vortices. Similarly, the angular momentum is Here n i a is the Killing vector corresponding to the specified rotation, a = 1, 2, 3, and we used Gauss law to obtain the right-hand side. We can now discuss the consequences of the vortex EFT for the CFT spectrum. To this aim, notice that the self-energy contribution ∼ log Q in eq. (2.14) is proportional to p q 2 p and implies that vortices with |q| > 1 are energetically unfavored 3 . The two main results of [30] are: • The lowest energy state for √ Q J ≤ Q consists of a vortex-antivortex pair rotating on the sphere, at a distance proportional to the spin ∆X/2 = J/Q (see fig. 1). The scaling dimension of the corresponding operator reads The leading correction to the ground state energy arises from the second term as a consequence of the logarithmic divergence of the vortex self-energy. This depends on the same coefficient α appearing in (2.14). The vortex mass contribution, given by the last term in (2.17), depends on a new coefficient and scales as the first subleading term in the ground state energy (2.4).
Corrections to this formula arise from the particle velocities and the phonon field. As J → √ Q, the vortices become relativistic and the derivative expansion breaks down.
• For Q J Q 3/2 the lowest energy state corresponds to a vortex crystal phase. Its energy is found approximating the vortex distribution as a continuous distribution and then minimizing the energy at fixed angular momentum. The result is For J √ Q there are no vortices and spinning operators are described by phonons (2.6).
3 Formulation of the EFT

Dual gauge field
As in 2 + 1 dimensions, to write a local coupling between vortices and the superfluid we consider a dual description in terms of a gauge field. Following the steps in sec. 2.2, we rewrite the leading order Lagrangian (2.1) in d = 3 using a two form Lagrange multiplier A µν = −A νµ : Integrating out v µ then gives Consequently, the homogeneous charge density j 0 = Q 2π 2 R 3 in the vacuum translates into a constant background field: The cutoff of the theory (2.3) is thus set by B 1/3 in the dual description. The action (3.2) is often called of Kalb-Ramond type and is invariant under the gauge transformations A µν → A µν + ∂ µ ξ ν − ∂ ν ξ µ , for an arbitrary vector ξ µ . The gauge redundancy allows imposing three gauge fixing conditions, since a gauge transformation generated by a total derivative ξ µ = ∂ µ α is trivial.
In the following, we shall be interested in fluctuations of the background (3.4). It is thus convenient to expand the gauge field in a background valueĀ µν plus fluctuations: where a possible choice isĀ Fluctuations are conveniently parametrized in terms of two three vectors b i and a i defined as: We partially fix the gauge requiring ∇ i A ik = 0, which sets the curl of b i to zero. Then the Lagrangian to quadratic order in the fluctuation reads: Following the gauge fixing, the field b i is purely longitudinal and corresponds to the phonon. Instead a i is a non-propagating degree of freedom, called the hydrophoton since the residual U (1) gauge invariance acts as a i → a i − ∂ i ξ 0 . Analogously to the Coulomb field in (2.9), the hydrophoton does not correspond to a local field in the original description and provides the leading coupling to the vortices.

String-vortex duality
Vortices in the dual description correspond to topological line defects, which are described as 1 + 1 dimensional strings embedded in the 3 + 1 dimensional spacetime [27]. The line element of a vortex p is parametrized by X µ p (τ, σ), where τ and σ are the world-sheet coordinates. We use the words "vortex" and "string" interchangeably. The Lagrangian is required to be Weyl invariant and reparametrization invariant for both τ and σ and is analogous to (2.11). The lowest order terms are given by The first term was discussed in the previous section. The second term is the leading coupling between a string of vorticity λ p ∈ Z and the gauge field. The last term is the generalized Nambu-Goto (NG) action for the vortex; in appendix C we derive its form via the coset construction. Here, the world-sheet metric is provided by: Since the superfluid velocity breaks Lorentz invariance, one can construct another independent symmetric world-sheet tensor, which can be chosen as In general the NG action contains an arbitrary function of G αβ h αβ , where G αβ is the inverse of G αβ . Weyl invariance further fixes the power of H which multiplies it. Finally dots in (3.10) stands for higher derivative terms.
Consider now the physical gauge X 0 p = τ for vortices. Using (3.6), the second term in (3.10) is linear in time derivatives of the vortex line. As we will self-consistently see in the next section, this implies that vortices move with drift velocity |˙ X| ∼ f /B ∼ B −1/3 . Then, similarly to what we argued below (2.11), terms of the kind˙ X ·˙ X in the NG action can be treated as higher derivatives and we neglect them. The coupling of the phonon field to the strings is also negligible to leading order. In this regime, the action reduces to where γ p = 6 1/3 F p (1) and we define Notice that the phonon spectrum (2.6) to leading order is not affected by the presence of vortices.
4 Results of the EFT

Classical analysis
From the leading order action (3.13) the following equations of motion for the hydrophoton and the strings are derived Eq. (4.1) is analogous to Ampère's circuital law in magnetostatic, a vortex acting as an electric current J i p sourcing the field f ij . Eq. (4.2) is the string equation of motion. Notice that it is first order in time derivatives and implies that vortices move with drift velocity |˙ X| ∼ f /B ∼ B −1/3 . The right-hand side arises from the NG action and it is proportional to the covariant derivative of the line element D Dσ ∂σX j (∂σX) ; the left-hand side comes from the minimal coupling to the gauge field. As in sec. 2.2 the electrostatic problem required the net charge on the sphere to be zero, the 3 + 1 dimensional magnetostatic problem defined by (4.1) and (4.2) requires zero vorticity flux on every closed surface. To this aim, we only consider closed strings.
The energy and angular momentum associated to solutions of the EOMs are computed from the stress energy tensor T µν = 2 √ g δS δg µν : The classical energy of the state is found from The first term is the energy of the homogenous ground state. The second term is the energy stored in the magnetostatic field f ij created by the vortices. Finally, the last term is the energy contribution from the tension and is proportional to the length of the string L p . Eq. (4.1) gives the field a i in terms of the string current: where G ij (x, x ) is the photon propagator on S 3 . In appendix A it is shown that the photon Green function on S d is where ∆X 2 is the chordal distance between two points in embedding space. Then the scaling dimension of the corresponding operator can be written as where α = 3 8π 2/3 c 1/3 . Notice the analogy with the structure of (2.14). The angular momentum (in units of 1/R) of the corresponding state is computed similarly: where n ab is the Killing vector corresponding to a rotation in the (X a , X b ) plane. Using Ampère's law (4.1) and Stoke's theorem, it is conveniently rewritten as where X p a are the vortex coordinates in the R 4 embedding of S 3 . The last equation on the righthand side is a formal notation for the area enclosed by the vortex projection in the (X c , X d ) plane. In the following we will study simple specific configurations.

Vortex rings
In nature, vortices often have a ring shape and move with a constant speed inversely proportional to the radius [26]. It is hence natural to look for vortex ring solutions of the EOMs (4.1) and (4.2). As we will see, a vortex ring generalizes the vortex-antivortex configuration in fig. 1. The simplest configuration one can study is a slowly moving vortex ring with unit negative charge λ = −1. We pick the gauge ξ = σ and consider a radius rR ≤ R ring in the (X 1 , X 2 ) plane in embedding space. The EOMs implies that the ring rotates with constant drift velocity v in the (X 3 , X 4 ) plane: 10) The precise value of v is fixed by eq. (4.1). From eq. (4.9) it follows that the only nonvanishing component of the angular momentum is given by: In figure 2 the motion is depicted in stereographic coordinates, defined by the relation (x, y, z) = 1 1+X 1 (X 3 , X 4 , X 2 ). Eq. (4.10) corresponds to a ring orbiting around the z axis; as the angular momentum is increased, the ring size increases and its velocity decreases. For r → 1 the surface embedded by the ring in the stereographic projection extends to cover the whole plane and the vortex lies statically on the geodesic corresponding to the z axis. Fig. 2 qualitatively generalizes the 2 + 1 dimensional motion depicted in fig. 1. Using (4.7) we can calculate the energy of this configuration as: The only nontrivial contribution arises from the second term, corresponding to the magnetostatic self-energy of the string. It diverges due to the short distance behaviour of the hydrophoton propagator. We regulate the calculation working in d + 1 spacetime dimensions and get Details of the computation are given in appendix B.1. There is a divergent piece for d → 3 proportional to the vortex length, which renormalizes the string tension. The contribution logarithmically enhanced by the cutoff ∼ e 2 r log r 2 B 2/3 can be seen as a consequence of the renormalization group running of γ [27] induced by the hydrophoton. Collecting everything, the scaling dimension (4.7) for a vortex ring state reads where we isolated the vortex contribution to the energy: Hereγ is a finite new coupling which absorbs all contributions proportional to r in (4.13). As in (2.17), the leading contribution arises because of the classical running of the tension induced by the magnetostatic self-energy and is given by the first term in (4.15). For J Q, the other contributions can be expanded in powers of the vortex length and to leading order effectively scale as Q 1/6 J 1/2 . Physically, this is understood noticing that the vortex energy density is set by e 2 ∼ Q 2/3 , hence for short vortices the energy can be estimated as the length times the energy density (neglecting the logarithmic running of the tension): 2πrR × e 2/3 ∼ Q 1/6 J 1/2 . However, as J → Q the functional dependence of the second and third term in eq. (4.15) deviates from this expectation, as a consequence of the vortex finite size. As the ring radius is decreased to inverse cutoff length r → 1/(ΛR), corresponding to J 34 → Q 1/3 , the magnetostatic field f ∼ e 2 /(Rr) becomes of the same order of the background field B and the vortex velocity approaches the relativistic regime. Hence subleading contributions to (3.13) become unsuppressed and the EFT breaks down. Eq. (4.14) can be identified as the minimal energy state at fixed angular momentum in its regime of validity.
We now study states with two vortices, one laying on the (X 1 , X 2 ) plane and the other on the (X 3 , X 4 ) plane in embedding space. Because of (4.9), these configurations are associated to operators in mixed symmetric representations of the SO(4) group. Consider first a radius R ring in the (X 1 , X 2 ) plane interacting with a ring of arbitrary size in the (X 3 , X 4 ) plane. In this geometry, the interaction does not affect the equations of motion and the solution takes a simple form vortex 1 : Focussing on negative unit charge vortices λ 1 = λ 2 = −1, this configuration corresponds to an operator in a mixed symmetric representation with spin given by Since the electric currents J i sourced by the strings are orthogonal, the corresponding scaling dimension is found analogously to (4.14): To leading order, a similar solution exists for 0 ≤ (1 − r 2 1 ) (RΛ) −2 , hence for 0 ≤ Q − J 34 Q 1/3 . As before, the consistency of the EFT requires J 12 Q 1/3 . In general, the mutual interaction affects non trivially the motion of the two vortex rings. One can, however, identify the logarithmically enhanced contributions analogous to the first term in (4.15) just from the free action. These indeed arise due to the tension running induced by the hydrophoton contribution to the vortex self-energy. For Q 1/3 J 12 , J 34 ≤ Q, the energy then reads: (4.19) This result holds as long as the minimal distance d between the two vortices is larger than the inverse of the cutoff:

Vortex crystals
Since the magnetostatic self-energy of a single vortex is proportional to λ 2 , strings with |λ| ≥ 1 are energetically unfavored. Hence the minimal energy state for values of the angular momentum J 34 Q is made by n 1 vortices. We then approximate the vortex distribution with a continuous current density J i (x). The corresponding state is found minimizing the energy (4.4) at fixed angular momentum (4.8), giving the following density profile: The leading contribution to the energy comes from the magnetostatic field and reads ∆ = αQ 4/3 + 3 4α Physically, this state corresponds to a vortex crystal [45,46,47]. When J 34 → Q 4/3 , the magnetic field f approaches B, vortices become relativistic and the EFT breaks down. Similarly, the ground state for Q J 34 , J 12 Q 4/3 is provided by a vortex crystal, whose current density and energy are given by

Quantization and Kelvin waves
Vortices in four dimensions are extended objects and can thus propagate Kelvin waves on them [27]. The corresponding states are then naturally associated to new operators of the CFT. To study them, we consider a single string of vorticity λ = −1. It is convenient to parametrize its coordinates via the following variables: w(t, σ) = X 3 (t, σ) + iX 4 (t, σ) = R cos η(t, σ)e iφ(t,σ) . (4.26) These are related through the constraint |z| 2 + |w| 2 = 1. We pick the gauge ξ = σ and t = τ . Integrating out explicitly the hydrophoton from eq. (3.13), we find the single vortex action as where F is given in (4.6). Eq. (4.27) can be seen as the (nonlocal) action of a complex field w(t, σ) living on R × S 1 . It is manifestly invariant under the action of the unbroken rotation generators J 34 , corresponding to rotations around the vortex w → e iα w, and J 12 , corresponding to translations along the string σ → σ + α.
We expand for small fluctuations around the background w = 0, which describes a radius R ring in the (X 1 , X 2 ) plane with J 34 = Q. The action to quadratic order in w reads: where S non−local is found expanding the second line in (4.27). It follows that the vortex is quantized as a standard non-relativistic field: [a n , a † m ] = 2πδ nm . (4.29) As usual the a n annihilate the vacuum a n |0 = 0, and thus so does w(t, σ). The proper frequencies ω n are computed in appendix B.2 and read Rω n ≡ ∆ k (n) = π(n 2 − 1) Notice that the n = 0 mode decreases the energy, while the n = ±1 modes have ω ±1 = 0. To understand this, it is useful to compute the angular momentum in terms of ladder operators to order O(Q 0 ). The rotations generated by J 12 and J 34 are linearly realized and their generators are quadratic in terms of ladder operators: The string realizes nonlinearly the full rotation group 4 . As a consequence, the broken components of the angular momentum are linear in the n = ±1 annihilation and creation operators:  As shown in fig. 3, a Kelvin wave in stereographic coordinates takes the form of a solenoid, trapping the magnetic field inside. The string undergoes a helical motion analogous to the one of a wine opener. Notice that Kelvin waves carry less energy than phonons with the same angular momentum (2.6). It follows that a state obtained acting on the vacuum as is the minimal energy state for the specified value of the angular momentum. This description applies in the linear regime m + 1 = Q − J 34 Q. When n = J 12 → Q 1/3 higher derivative terms become unsuppressed and the EFT breaks.

Higher order corrections
Corrections arise from higher derivative terms we neglected in (3.13) and are suppressed by powers of the cutoff scale (2.3). Following [30], here we comment on their form.
The first class of corrections was discussed in [10,13] and arises considering the effect of curvature terms in the superfluid and vortex action; these corrections are controlled by the sphere radius and hence scale as 1/(ΛR) 2 ∼ 1/Q 2/3 . These are present also in the absence of vortices.
Focus now on the single vortex state described in sec. 4.2. We find corrections controlled by the vortex length L ∼ J 34 /Q, which hence scale as 1/(ΛL) 2 ∼ Q 1/3 /J 34 . They arise from the terms we neglected in the NG action to write (3.10) and are proportional to (∇ i b i )/B, f 2 /B 2 and˙ X 2 . Higher derivatives of the string line element as well as the phonon contribution to the energy (3.2) belong to the same class. Similarly, there are corrections of the form Q 1/3 /J 34 , Q 1/3 /J 12 to eq. (4.18) for a two vortex state. Notice that the subleading Q 2/3 term in the ground state energy is bigger than the vortex contribution (4.15) for Q 1/3 J 34 Q. The latter gives instead the leading contribution for J 34 ∼ Q. The vortex contribution is anyway functionally distinguished from the ground state energy correction and is thus always calculable.
Let us now turn our attention to the Kelvin waves discussed in 4.4. The same corrections discussed for a vortex ring exists in this case. Furthermore, for n 1 higher derivative corrections to the single vortex action (4.28) become important. As typical for a non-relativistic field, these arise due to terms with two time derivatives, or, equivalently, with four space derivatives (suppressed by an extra H −2/3 factor by Weyl invariance) and scale as n 2 /Q 2/3 = J 2 12 /Q 2/3 . Notice that the relative corrections to the ground state energy of the vortex are bigger than the Kelvin wave energy (4.30) for J 2 12 /Q 1/3 Q 1/2 /J 1/2 34 ; however, these corrections are independent of J 12 , which enters only through (4.30).
Finally, the leading corrections to the energy of the vortex crystals states discussed in 4.3 arise both from the phonon contribution to the energy, which is proportional to (∇ i b i ) 2 /f 2 ∼ J ab /Q 4/3 2 , and from the free tension contribution, which gives Q/J corrections using (4.21) or (4.23). Here J ab stands for both J 34 and/or J 12 depending on the state.

Correlators
We now turn our attention to the study of correlators. As in [30], the most natural correlation function 6 which can be studied corresponds to a current insertion within two equal vortex states. In the EFT, this is determined through the following relations: The hydrophoton field is obtained from (4.1), which, following the analogy with Ampère's law, can be conveniently rewritten in integral form as where λ enc is the vorticity flux through the surface enclosed by the curve C. Using this, eq.s (5.1) can be used to make nontrivial predictions about the OPE coefficients of the theory. Consider first the traceless symmetric state corresponding to a radius R vortex in the (X 1 , X 2 ) plane, which has J 34 = Q and J 12 = 0. For this state, eq. (5.1) reads: The expectation value of a spin-1 parity even conserved operator in a traceless symmetric state |(J, J), J 34 = 2J, J 12 = 0 is [48]:    Cutting off the sums at m √ J , we obtain the following predictions Analogously, for the state (4.24) with Q J 12 , J 34 Q 4/3 , the EFT gives

Vortices in arbitrary dimensions
Based on the considerations so far, as well as on the previous results of [30], it is not hard to understand the qualitative feature of the vortex EFT in generic dimensions. We give some brief comments here for completeness. We focus on the derivation of the scaling dimensions for traceless symmetric operators in generic higher dimensions. We first need to construct the dual of the d One can similarly write the Nambu-Goto like action for the membrane [51]; we do not report here the expression since its detailed form will not be needed in the following.
One can now proceed as in sec. 4. From the energy momentum tensor, one finds that the leading contribution to the vortex energy comes from the hydrophoton gauge field. Generalizing eq.s (2.16) and (4.9), the angular momentum is proportional to the volume enclosed by the vortex in embedding coordinates. For Q 1/d J ≤ Q, the minimal energy state corresponds to a single spherical vortex. The leading contribution to the vortex energy arises from the running of the tension, induced by the hydrophoton contribution to the self-energy as in (4.13). This can be computed using a flat space approximation for the gauge field Green function and a UV hard cutoff Λ ∼ Q 1/d /R to regulate the result: where ∆ 0 (Q) is given by (2.4). We expect d dependent corrections of order J 3), similarly to (4.14); these contributions however will not be logarithmically enhanced by the cutoff. As in section 4.3, for Q J Q d+1 d we can identify the minimal energy state as a vortex crystal. Following the same steps which lead to (4.22), we find that the energy of this state is Eq.s (6.3) and (6.4) match the results obtained in [30] and in this paper for d = 2, 3.

Conclusions and future directions
Condensed matter phases often admit a simple effective description [22]. In CFTs, one can take advantage of this using the state/operator correspondence to study CFT data at large quantum numbers. In this work, these ideas were used to compute the scaling dimensions of operators of large internal charge and spin in a U (1) invariant CFT 4 . The results obtained for traceless symmetric operators can be seen as a generalization of those obtained in CFT 3 [30]; however the study of operators in mixed symmetric representations explored qualitatively distinct regimes, such as Kelvin wave propagation on a string (sec. 4.4). We also provided predictions for correlators of the U (1) current in between vortex states in sec. 5 and generalized the predictions for the scaling dimensions of traceless symmetric operators to arbitrary dimensions in eq.s (6.3) and (6.4). The most direct extension of this work would be a detailed analysis of higher order corrections, both in three and four dimensions. In particular, a refinement of the continuum approximation used in sec. 4.3 might allow for the study of collective excitations in the vortex crystal phase, corresponding to new CFT operators, possibly similar to the Tkachenko mode studied in [46,47].
Within the exploration of the superfluid phase of CFTs [10,13,14,15,16,17,18,19,20], the non-Abelian case still leaves some open questions. As argued in [13], the basic prediction for the scaling dimension of the lightest charged operator is insensitive to the non-Abelian nature of the symmetry group, which instead manifests itself via the existence of the so-called gapped Goldstones [52,53]. Massive Goldstones are crucially needed to close the current algebra of the non-Abelian group, like standard Goldstones, but at the same time have a fixed gap of order cutoff dictated by the symmetry. Their role in the large charge sector of CFTs and in more general finite density QFTs deserves further investigation [54].
Most of the existing results for large charge operators in CFTs are derived under the assumption that the CFT admits a superfluid phase. It is hence important to check whether this assumption applies in known theories. So far, most computations focussed on the prediction for the scaling dimension of the lightest charged operator. This has been verified in Monte-Carlo simulations of the O(2) [55] and O(4) [56] model and perturbatively for Monopole operators [59,60,61,62,63], to order O(N 0 ) in the CP N model [57] and at leading order in the number of flavors in QED 3 and the gauged Gross-Neveu model [58]. Relatedly, large charge states have been studied in AdS/CFT in the context of holographic superconductors [64,65,66]. We are currently addressing similar questions within the ε-expansion [67], which allows for extensive checks of the EFT predictions. Perhaps, the techniques explored in these works might also find application in a different context, such as the study of processes with many external legs within the Standard Model [68,69,70,71].
Despite their generality, superfluids are not the only possible description for large charge states in CFT. For instance, if one assumes the charge Q to be unbroken, a natural phase 7 which might describe the CFT is a Fermi liquid [74]. Different descriptions are also possible and are expected to apply in the presence of moduli spaces, which naturally allow for light degrees of freedom other than the Goldstone mode. This is the case for free massless theories and N = 2 superconformal field theories, where the large R-charge expansion is organized differently [75,76,77]. In [78,79] the possibility of a semiclassical but inhomogeneous phase in the O(4) model was also explored.
In [80], the question of how to find solutions to the crossing equations at large charge was addressed in connection with the existence of a macroscopic limit of correlators [81]. However, it remains an open question whether one can relate explicitly the large charge sector of CFTs with the spectrum of light operators. Perhaps relatedly, a bootstrap analysis recently connected large scaling dimensions tails of weighted spectral density of primary operators with light operators exchanged in the dual channel [82].
Finally, the predictions of analyticity and the existence of a perturbative expansion for large internal quantum numbers are reminiscent of the bootstrap results for large spin operators [7,8,31,32,33,34,35,36,37,38,39]. The physical picture behind those results is particularly clear in the dual AdS space 8 [8,31], where double-trace operators are associated to two widely separated objects. Because of the AdS geometry, these only interact weakly via the exchange of highly off-shell modes. A universal EFT description might exist in this case as well.
Consider the action of a massless vector field coupled to a conserved current J µ (in Euclidean signature): The gauge field on the equations of motion is given by where G µν (x, x ) satisfies the equation Here Λ ν is a pure gauge term which drops from physical observables; primed and unmprimed indices refer, respectively, to the points x and x . Let us define the following biscalar where (X − X ) 2 is the chordal distance in embedding coordinates. Given the isometries of the sphere, it is possible to parametrize the propagator as The last term is gauge dependent and drops from eq. (A.2). The following properties hold: These can be easily explicitly verified in stereographic coordinates, for instance. It follows By symmetry, we can write Λ ν (x, x ) = (∂ ν u)Λ(u). Then for x = x (A.3) gives two equations: We can integrate the second and plug the result in the first to obtain This is just Klein Gordon equation for a scalar field of mass m 2 = d − 2 on S d . The solution is fixed requiring a power low singularity for u → 0 and regularity at the antipodal point u → 2 [87]: The normalization is determined matching the short distance limit with a flat space propagator. Plugging in (A.5), we get eq. (4.6) in the main text.

B Vortex energy in dimensional regularization
To regulate the computation of the magnetostatic energy, it is convenient to work in d+1 spacetime dimensions. It is natural to modify the Lagrangian (3.2) in a way which preserves Weyl invariance: The definition of H in terms of the two-form field A µν here is unchanged. Notice that working in arbitrary d with a 2-form field we loose the duality with a shift invariant scalar, hence this regularization breaks U (1) invariance at intermediate steps in the calculation 10 . A more responsible approach might be, perhaps, to promote A µν to a d − 2 form field, preserving Weyl invariance of the action. An investigation of this issue might be helpful in expanding the result of this paper to subleading orders. Expanding the action (B.1) to quadratic order gives where we defined the electric coupling in d space dimensions as The NG action discussed in section 3.2 is unchanged in d dimensions.

B.1 Vortex ring self-energy
Consider a single vortex moving on a trajectory given by (4.10). We want to compute the selfenergy contribution due to the hydrophoton, i.e. the second term in eq. (4.12). In Hopf coordinates (1.13) and in dimensional regularization, it reads: where we isolated the integral (B.5) 10 Conversely, a cutoff approach as in [30] breaks Weyl invariance.
In d = 3, the integral is logarithmically divergent for ξ → 0, corresponding to the interaction of an infinitesimal line element with itself. Setting 1 2 (1 − cos ξ) = y in (B.5), we get The divergent part comes from the first term in the expansion of the hypergeometric function when the argument goes to one: We separate explicitly this contribution and recast the integral as: where the divergent piece is and the regular part can be evaluated directly in d = 3, where it reads To compute the latter, it is convenient to use the following expansion Interchanging sum and integral, the regular part gives Collecting everything and adding the tension contribution, we arrive at (4.13).

B.2 Kelvin waves frequency
The EOMs which derive from (4.28) give the oscillation frequency of Kelvin waves as: where the second term comes from the nonlocal piece of the action and is written in terms of the following integral: δω I n = 1 2 dσ n 2 cos(nσ) − cos σ F (1 − cos σ) + cos 2 σ − cos(nσ) cos σ F (1 − cos σ) . (B.14) Let us sketch the evaluation of (B.14). Changing variables as before, we write δω I n as the sum of the following two contributions: (B.16) Here T n (x) = cos (n arccos(x)) is a Chebyshev polynomial. The divergent contributions are identified from the leading term of the Hypergeometric expansion (B.7) and can be evaluated using (B.17) To evaluate the regular parts, we use the following results: Using Mathematica we computed these integrals explicitly for fixed integer values of n and identified their functional form; the result was then verified numerically and using the n → ∞ asymptotic expansion of the results (B.18) and (B.19). This indeed can be obtained explicitly truncating the series expansion of the Hypergeometric functions in the integrals and using (B.17). The regular contributions from I 1 (n) and I 2 (n) finally read The second contribution in (B.21) arises since we subtracted the O 1/ √ y term in the expansion of the Hypergeometric function from the first piece, in order to use (B.19). Collecting everything and expanding for d → 3, we find the following remarkably simple result: .
C Nambu-Goto action from the coset construction In [13], it was argued that two charged scalar operators insertions in d + 1 dimensions at x = 0 and x = ∞ induce a specific symmetry breaking pattern for the leading trajectory in the path integral. A similar logic can be applied when the operators have also large spin J. As in the scalar case, translations P µ , special conformal transformations K µ and dilatation D are broken, with the combination D + µQ left unbroken. Assuming the operator insertion to be polarized in the (x 1 , x 2 ) plane, the Lorentz generators J 1p , J 2p with p, q = 0, 3, . . . must necessarily be broken. A vortex corresponds to the regime where it is energetically favorable for the system to still be in an almost homogeneous state, rotations being broken by a localized region of size 1/j 0 ∼ R/Q 1/d in which the superfluid description breaks. This region naturally extends from 0 to ∞ along the directions orthogonal to the spin polarization, corresponding hence to a d − 1 dimensional membrane. In this regime, J 12 parametrizes rotation around the vortex and it is thus unbroken. We then identify the symmetry breaking pattern corresponding to a vortex as: where we introduced the set of indices m, n = 1, 2 and p, q = 0, 3, . . . . In order to apply the coset construction in a curved manifold, it is convenient to think in terms of the generators acting in a local chart, denoted { D, P µ , K µ , J µν }. These are naturally associated with those acting on the plane considering the formal R → ∞ limit on R × S d [13]: We then rewrite the symmetry breaking pattern (C.1) in terms of the hatted generators. Focussing on 2 + 1 and 3 + 1 dimensions, we get 2 + 1 :    P 0 = P 0 + µ Q, J 12 unbroken, P i , J 0i , K µ , Q broken; 3 + 1 :    P p = P p + µ Q, J 12 unbroken, P m , J 0i , J n3 , K µ , Q broken.

(C.3)
From (C.3) we can construct the Nambu-Goto action for the vortex via the coset construction [88,89], applied to the case of a membrane [51].